If only a small fraction of the chip layers get multiple patterning or double patterning, and most design rules on the SOC are very loose, the extra costs will be diluted. So the "worst case" scenario shouldn't be so bad.
I think the concern is that with the expense of the fancy litho tools, most fabs would only have 1 or 2 initially. Then when you have all of the critical layers needing to go through the tool multiple times, (active, poly, contact, metals and vias) you end up with wafers just queued behind the litho tool all_the_time - because you don't just have 1 lot at a time running in your fab.
A healthy mix of technologies will be required. Who will make the masks for Imprint technology or inspect and repair them 1x!!
Who will pattern the base structures for directed self assembly, repetitive patterns only??
It might work for memory cells, others?
Rapid prototyping and critical layers of the 1xnm and 2xnm nodes, E-Beam dirct write will be the solution.....no masks, easy to change, or simulate process changes, adapt depending on the flow changes across the wafer..
A lot of challenges ahead....smart device integration might be a better way to improve the performance of devices, not just scaling!!
Truly a shame that more 'professionals' don't look more seriously at e-beam lithography themselves, and not take the word of people in the photomask industry, who would directly be threatened if mask-less lithography came into popularity. E-beam litho has demonstrated nano-level capability for more than 20 years, with comparative overlay capability equaling the best aligners out there. Yes, throughput is an issue, but at 1/6 (or less) the cost of an good immersion stepper, you can buy several!!! Truth is that even the finest nano-imprint tool or EUV stepper will require masks, and they will only be available via e-beam lithography (as well as being ridiculously expensive and short lived). Direct write is a viable technology - today!
E-beam is more mature than optical for sure. But electrons, primary, photo-, or secondary, are prone to random disturbance. And they go into the substrate. So the interest in DSA, though it seems sensitive to the guiding pattern size.
The dream of e-beam direct write has been around as long as x-ray lithography, and just as successful. The e-beam problems of throughput, data management and error correction simply cannot be solved in time, or economically. EUV, or, more correctly, soft x-ray projection lithography, continues to suffer from very x-ray like problems of decades ago. The only technology with the potential to complement optical lithography is imprint, which essentially is optical lithography: it uses an I-line source, I-line resists, and quartz based photomasks. Defects are a more manageable challenge than those facing EBDW and SXPL (EUV), particularly in memory. Lithography will bifurcate into solutions for logic and solutions for memory.
Perhaps, regarding as the viewpoint of mass-production, the approach to improve the common DOF of the different characteristics of pattern,ex. iso- vs. dense- ,or line vs. space, is more practical and economic than that to put all resource to enhance the resolution.
Blog Doing Math in FPGAs Tom Burke 2 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...